Optical fibers are generally accepted as the transmission medium for most long distance lightwave communication systems. Theoretically, because of almost unlimited bandwidth, fibers can transport thousands of very high speed data channels simultaneously. Realistically, however, loss, dispersion, and nonlinear intensity-dependent effects combine to determine both the number of channels supported by the fiber and the spectral placement of the channels on the fiber. In an effort to reach a maximum data throughput rate, most communication systems are designed as wavelength-division multiplexed (WDM) or frequency-division-multiplexed (FDM) systems having carrier signals transmitted at and in close proximity to a benchmark transmission characteristic for the optical fiber, namely, the zero dispersion wavelength, .lambda..sub.0. This wavelength is defined as the point at which the second derivative of the propagation constant for the fiber taken with respect to wavelength is zero. Standard single-mode fibers exhibit zero dispersion nominally at 1.3 .mu.m, while dispersion-shifted fibers have a zero dispersion point at approximately 1.5 .mu.m.
While it has remained a foregone conclusion that one carrier in a WDM or FDM system be transmitted at the zero dispersion wavelength, the placement of other carriers for adjacent channels has been a subject of important study. In fact, several researchers have modeled the problem of carrier placement as a three-wave or four-wave mixing problem to account for nonlinear effects arising when intense lightwave signals propagate in the optical fiber. See, K. O. Hill et al., J. Appl. Phys., 49(10), pp. 5098-5106 (1978) and N. Shibata et al.,IEEE J. Quant. Elect., Vol. Qe-23, No. 7, pp.1205-1210 (1987). Both Hill et al. and Shibata et al. describe processes in which three input lightwave signals with different frequencies generate lightwave signals at as many as nine corresponding new frequencies. The new lightwave signals result from frequency mixing or crosstalk. Shibata et al. conclude by extrapolating their results from experiments at short wavelengths that, in a frequency multiplexed transmission system with one narrow linewidth carrier signal at the zero chromatic dispersion wavelength of the fiber, transmission of adjacent lightwave signals should occur with frequency separations greater than 400 GHz (2.25 nm) at .lambda..sub.0 =1.3 .mu.m and greater than 300 GHz (2.4 nm) at .lambda..sub.0 =1.5 .mu.m to achieve complete suppression of lightwave signals generated through four-wave mixing. It is now understood by me that the conclusion by Shibata et al. is flawed and that complete suppression does not occur in the frequency multiplexed system described by Shibata et al.